การจำแนกบริเวณเลือดในทางเดินอาหารจากภาพถ่ายกล้องแคปซูลไร้สายด้วยวิธีการ K-Means Clustering, วิธีการK-nearest neighbor และวิธีการจัดกลุ่มข้อมูลด้วยค่าสี RGB

Abstract

Wireless capsule endoscopy has now been used in medicine to take pictures of the gastrointestinal tract. The wireless capsule camera takes many pictures, which makes it difficult to analyze with the naked eye. Therefore, in this paper, three methods of automatic classification of bleeding areas were proposed: K-Mean Clustering, consisting of before color Thresholder and after Color Thresholder, KNN, and Clustering methods. RGB color values were used to classify the bleeding area, and a confusion matrix was used to test the efficiency. The results of the system consisted of the before color Thresholder, after Color Thresholder, KNN, and grouping data by RGB color values, It was found that the loU was used to measure the performance , each of which scored 51.18%, 59.64%, 43.06%, and 61.88%, respectively. When changing the image dataset, the results of the system consisted of the before color Thresholder, after Color Thresholder, and grouping data method by RGB color values as follows, loU scored 15.77%, 19.77%, and 46.22%, respectively. Based on the performance experiments conducted on the two data groups, it was demonstrated that the RGB color data grouping method was the most efficient. As it pertains to a specific classification of blood regions, this method's RGB values encompass a more comprehensive range of color values than other methods.